Comms Agent · Phase 1 Operations Guide

For: [ESCALATION_OWNER] (Primary), [BACKUP_OPERATOR] (Backup) Coverage Period: February 5 to March 5, 2026 ([OWNER]'s leave) Version: 1.3 · Last Updated: February 1, 2026

Quick Reference Card

1. What the Agent Does Automatically

Low Touch Requests (55% of volume, design target)

The agent handles these without human intervention:

1. Receives form submission
2. Classifies as Low Touch
3. Retrieves draft content from linked document
4. Revises applying Smart Brevity rules
5. Creates Email Draft Doc with rationale
6. Emails requester with link to their draft
7. Logs to tracking sheet
8. Records engagement summary

You'll see: Nothing (unless there's an error)

Requester sees: Email with link to their revised draft

What Triggers Auto-Escalation

Even if a request looks simple, the agent escalates to you when:

2. What You Handle

When Escalation Arrives

You'll receive:

1. Email from "[TEAM] Communications" with a HIGH or MEDIUM prefix
2. Chat Space ping in the team comms space
3. Escalation Starter Doc (link in email)

Your Workflow

1. Open Escalation Starter Doc
2. Review request context and classification reasoning
3. If audience is an unfamiliar senior IC, run engagement summary (Section 3)
4. Draft content using outline provided
5. Apply audience-specific guidance from Starter Doc
6. Send draft to requester for feedback
7. Iterate and deliver
8. Update tracking sheet: Status -> "Completed"

Escalation Starter Doc Contents

Every Medium/High Touch escalation creates a doc containing:

3. Running Engagement Summaries (Manual)

When to Run One

Run an engagement summary when:

• Target audience is a specific senior IC you haven't worked with
• Request involves sensitive communication to leadership
• Requester mentions needing to "land the message" with someone specific
• Starter Doc flags "Engagement summary recommended: true"

How to Run

For an individual team member:

1. Install [internal browser extension] if not already
2. Navigate to the person's [internal directory] profile
3. Pin the page in the extension
4. Open a new chat and paste this prompt:

Analyze this team member for a communications engagement strategy.

Search across their profile, recent docs, presentations, and any visible org announcements. Focus on:

1. **Role & Strategic Position**: Title, team, what they're accountable for
2. **Communication Preferences**: How do they write? Async vs sync? Detail level?
3. **Current Priorities**: What are they working on now? What do they care about?
4. **Strategic Hooks**: What framing would resonate given their priorities?
5. **Tactical Do's and Don'ts**: Based on their style, what works/doesn't?

Output format:
- Header: Name, Role, Level
- Executive Engagement Strategy (2-3 sentences)
- Strategic Hooks (bullet list)
- Communication Intelligence (preferences, style)
- Tactical Do's and Don'ts (table format)

Exclude any artifacts older than January 2023.

1. Review the output and use the insights when drafting the communication.

For a group/alias:

1. Navigate to the distribution list page for the alias
2. Pin the page in the extension
3. Paste this prompt:

Analyze this distribution list for a communications engagement strategy.

Search the group membership, recent threads, and any visible documentation about the group's purpose. Focus on:

1. **Group Mandate**: What is this group for? Who are the key stakeholders?
2. **Collective Priorities**: What topics dominate recent discussions?
3. **Communication Norms**: Formal vs informal? Pre-reads expected? How do decisions happen?
4. **Key Personas**: Who are the vocal/influential members?
5. **Strategic Do's and Don'ts**: What works for this group?

Output format:
- Header: Group Name, Mandate
- Group Engagement Strategy (2-3 sentences)
- Key Stakeholder Personas (3-5 archetypes)
- Communication Intelligence
- Strategic Do's and Don'ts (table format)

Exclude artifacts older than 18 months.

Where to Save Engagement Summaries

Save to: [OUTPUT_FOLDER] / Client Engagement Summaries / [email_or_group].md

The agent checks this folder and will use it for future interactions.

4. [VP_NAME] Communications (HIGH TOUCH)

All [VP_NAME]-involved requests are HIGH TOUCH. The Starter Doc includes a [VP_NAME] prompt for the executive-voice assistant.

Workflow for [VP_NAME] Comms

1. Open Starter Doc
2. Scroll to the [VP_NAME] prompt section
3. Copy the entire prompt
4. Open the [VP_NAME] executive-voice assistant
5. Paste the prompt
6. Review output against recent [VP_NAME] communications for voice consistency
7. Before sending to [VP_NAME]: review with [COMMS_REVIEWER]
8. Iterate based on [VP_NAME]'s feedback
9. Coordinate timing and distribution

[VP_NAME]'s Voice (Quick Reference)

• Direct but warm
• Data-driven with human context
• Strategic framing connecting to bigger picture
• Acknowledges complexity without hedging
• Celebrates wins without being boastful
• Forward-looking and action-oriented

5. Common Scenarios

Scenario A: Draft Access Error

You receive: Email saying agent couldn't access requester's draft What happened: Requester linked a doc the agent's service account can't open Action:

1. Requester already received email explaining the issue
2. Wait for them to fix sharing and resubmit, OR
3. Ask them to copy/paste content directly to you

Scenario B: Site/Website Request

You receive: MEDIUM TOUCH escalation for website content What happened: Agent detected site-related keywords and escalated (can't make IA decisions) Action:

1. Review Starter Doc for site-specific outline
2. Assess where content fits in existing site structure
3. Draft content following site style guide
4. Coordinate with relevant site owner before publishing

Scenario C: Unclear Request

You receive: MEDIUM TOUCH with "Confidence: 65%" What happened: Agent wasn't sure how to classify Action:

1. Review Starter Doc for what confused the agent
2. Reply to requester asking clarifying questions
3. Once clear, proceed with appropriate workflow

Scenario D: Request Bypasses Form

You receive: Chat ping saying someone emailed the ops alias directly Action:

1. Reply to sender: "Thanks for reaching out! To help us triage effectively, please submit via [intake form]"
2. If urgent, process manually and encourage form use next time

6. Tracking Sheet Management

Key Columns

Your Updates

When you complete an escalated request:

1. Change Status to "Completed"
2. Add final doc link if different from Starter Doc
3. Add notes in Additional Notes column if needed

7. Error Handling

Agent Error Ping

You receive: AGENT ERROR in Chat Space Action:

1. Note the timestamp
2. Go to the script project, then View, then Executions
3. Find the failed execution and view logs
4. Common fixes:

• Model API error: Usually rate limit, will self-resolve
• Sheet error: Check if columns shifted
• Permission error: Check folder/sheet sharing

1. If unclear, note the error and escalate to [BACKUP_ENGINEER]

Manual Reprocessing

If the agent missed a submission:

1. Find the row in the tracking sheet
2. Note the form data
3. Process manually using the same workflow

8. Contact Escalation

9. Weekly Rhythm

10. Metrics to Review (For [OWNER]'s Return)

When [OWNER] returns from leave, these metrics will help assess agent performance and identify improvements.

Key Metrics to Track

Quality Signals to Note

What [OWNER] Will Do With This

1. Tune triage criteria based on misclassifications
2. Expand knowledge base: new patterns, examples, audience profiles
3. Optimize automation rate: identify requests that could be Low Touch with better prompting
4. ROI calculation: compare time saved vs. manual processing

Appendix: Agent Test Functions

If you need to verify the agent is working:

1. Open the script project
2. Run testGeminiConnection(): should log "Success"
3. Run testChatSpaceWebhook(): should ping the Chat Space

If tests fail, check Script Properties are still configured.

Questions during [OWNER]'s leave? This guide should cover 95% of scenarios. For edge cases, use your judgment.

Back to the writing room